Fan tower and blade self-adaptive protection system and method based on visual monitoring

Abstract

The present application relates to offshore wind power safety monitoring technology, and specifically discloses a wind turbine tower and blade adaptive protection system and method based on visual monitoring. The present application dynamically constructs a horizontal reference resistant to wave interference through an onboard high-definition camera capturing sea surface flare, and fuses tower edge recognition to calculate the inclination in real time. When the inclination continuously exceeds the limit, the blade is segmented to contract to reduce wind load, and the inclination change rate after contraction is used to distinguish between elastic deformation and structural failure. For the case where the inclination does not change, a coaxial laser scanning device is started to excite local vibration of the tower, and the loosening point is located by analyzing the time delay of vibration wave conduction. The present application can realize millimeter-level precision monitoring of tower inclination; the active protection mechanism of the blade reduces the risk of accidents; improves the precision of non-contact fault location, and reduces the diving maintenance operation. It is mainly suitable for structural health monitoring and safety protection of large offshore wind turbine generators.

Description

TECHNICAL FIELD

[0001] The technical field of offshore wind turbine structure health monitoring and safety protection and the technical field of fault diagnosis, in particular to a wind turbine tower and blade adaptive protection system and method based on visual monitoring, especially suitable for real-time monitoring and intelligent diagnosis of equipment state in high temperature, high humidity and strong radiation environment. BACKGROUND

[0002] The large-scale development of offshore wind turbines leads to an exponential increase in tower moment load. The existing inclination monitoring technology faces three bottlenecks: the traditional magnetic attraction type inclination sensor (such as the technical solution disclosed in the online monitoring device for inclination of offshore wind power foundation disclosed in the publication No. CN221859553U, and the patent name is a kind of online monitoring device for inclination of offshore wind power foundation) has an annual failure rate of 43% in a salt spray environment and an accuracy of only ±0.1°, the welding installation scheme (such as the technical solution disclosed in the offshore wind turbine tower sensor device, vibration detection system and inclination detection system disclosed in the publication No. CN209129789U, and the patent name is an offshore wind turbine tower sensor device, vibration detection system and inclination detection system) causes the local stress concentration coefficient of the tower to exceed 2.5 and accelerates the initiation of fatigue cracks; the Bayesian early warning model (such as the technical solution disclosed in the publication No. CN118708980A, and the patent name is a kind of offshore wind tower inclination deformation prediction method and system) causes a response delay of up to 126 minutes due to the lack of an actuator, and the wave disturbance causes the inclination data to oscillate by ±0.3° (the response delay of the floating platform liquid bag monitoring device is more than 5 seconds), the displacement conversion error of the acceleration sensor is as high as ±15% and cannot represent the overall bending mode.

[0003] Therefore, it is urgent to develop a non-contact, high-precision and active protection capability inclination monitoring closed-loop system. SUMMARY

[0004] In view of the above problems existing in the prior art, the purpose of the present application is to provide a wind turbine tower and blade adaptive protection system and method based on visual monitoring.

[0005] The present application provides the following technical solutions:

[0006] The wind turbine tower and blade adaptive protection system based on visual monitoring comprises a visual monitoring device, a blade length adjusting unit arranged on the blade, and a fault response unit;

[0007] The visual monitoring device comprises a vibration analysis module and a salt spray resistant sealed cabin arranged at the bottom of the wind turbine cabin, the vibration analysis module is used for analyzing the displacement field of the vibration wave propagation image, calculating the vibration wave arrival time difference on both sides of the weld, comparing the real-time waveform with the standard waveform library, identifying the conduction delay anomaly, and the salt spray resistant sealed cabin is integrated with a camera, a pulse laser and a scanning galvanometer;

[0008] The blade length adjusting unit comprises at least one telescopic carbon fiber blade segment for adjusting the blade length.

[0009] The fault response unit is used to activate the bolt pre-tightening device at the corresponding position when locating the loosening point, and to start the cement grouting system when the vibration frequency deviation is greater than the set threshold.

[0010] The protection method of the fan tower and blade adaptive protection system based on visual monitoring includes the following steps:

[0011] S1, through the camera installed at the bottom of the fan cabin, continuously shoot the video containing the tower and the sea surface at the set pitch angle;

[0012] S2, dynamically construct the absolute horizontal reference line of the sea surface;

[0013] S3, extract the edge lines of the tower on both sides, and calculate the average angle between them and the horizontal reference line as the real-time inclination θ;

[0014] S4, when θ is greater than the set angle threshold and lasts for t seconds, start the blade length adjusting unit to shorten the blade length in sections;

[0015] S5, monitor the inclination change rate after the blade is shortened: if the change rate is greater than the set threshold A, it is determined that the elastic deformation is maintained; if the change rate is less than or equal to the set threshold B, start the laser scanning diagnosis;

[0016] S6, move the laser focus of the pulse laser controlled by the scanning galvanometer along the tower weld, and synchronize the analysis of the vibration wave propagation image captured by the camera to locate the loosening point coordinates.

[0017] Further, the specific process of S2 is as follows:

[0018] Detect the solar flare belt profile of the sea surface, fit it into a smooth reference line after wavelet denoising, and fuse the IMU data to compensate for the fan swing.

[0019] Further, the specific process of S3 is as follows:

[0020] S3.1, extract the edge lines of the tower on both sides through the edge detection algorithm;

[0021] S3.2, construct a dynamic horizontal reference line;

[0022] S3.3, calculate the inclination deviation ;

[0023] S3.4, optimize the output real-time inclination θ based on the deep reinforcement learning model;

[0024] S3.5, update the system state in combination with the wind speed and blade length.

[0025] Further, the specific process of S3.1 is as follows:

[0026] 3.1.1) input the original image;

[0027] 3.1.2) suppress sea wave, salt spray noise in the original image by using Gaussian filter;

[0028] 3.1.3) calculate the derivative of the original image in X and Y directions respectively using sobel operator, and calculate the gradient intensity and direction of each pixel point based on the obtained derivative;

[0029] 3.1.4) traverse each pixel point, check its two adjacent pixels in the gradient direction, if the gradient intensity of the current point is greater than the intensity of the two adjacent points, keep the point as a candidate edge point; otherwise, set the suppression gray value of the current point to 0;

[0030] 3.1.5) calculate strong and weak edges to distinguish the true edges and the false edges caused by noise;

[0031] 3.1.6) after processing based on the edge detection algorithm of steps 3.1.1)-3.1.5), the original image is converted into a binary black and white image, wherein the white pixel points constitute all the edges, and based on the Hough transform algorithm, straight line segments are detected from the white points, and from all the detected straight lines, two straight lines (close to vertical, long in length, and symmetrically appearing on the left and right sides in the image) that meet the tower drum edge characteristics are found, that is, the edge lines.

[0032] Further, the specific process of S3.2 is as follows:

[0033] The wavelet denoising algorithm is adopted to extract the sea wave mirror reflection profile, the Kalman filter is adopted to suppress wave interference, and a sub-pixel level stable reference line is formed; at the same time, the tower drum area is segmented by fusing LiDAR point cloud data, and the angle between the tower drum axis and the reference line is calculated in real time by the Canny edge detection algorithm.

[0034] Further, in S3.3, the formula for calculating the tower drum inclination deviation is:

[0035] ;

[0036] wherein, is the tower drum inclination deviation, is the real-time measured tower drum inclination, is the theoretical tower drum inclination corresponding to the dynamic horizontal reference.

[0037] Further, in S3.4, the inclination adaptive prediction and control algorithm based on deep reinforcement learning is adopted, and the calculation formula is:

[0038] ;

[0039] ;

[0040] ;

[0041] wherein, represents the Q value of taking action under state ; is a learning rate, which is adaptively adjusted according to the inclination angle change to balance the stability and convergence speed of the algorithm; is the reward value at the current moment; gamma is a discount factor; is a reference inclination angle change rate.

[0042] Further, in the S4, the calculation formula of the blade segment shortening amount is:

[0043] ;

[0044] wherein, ΔL is the length of each blade shortening, K is a stiffness attenuation coefficient, which is obtained by exponential fitting of historical inclination angle data, θ is the real-time inclination angle, θ0 is the initial inclination angle, V is the current wind speed, and V0 is the rated wind speed.

[0045] Further, it further comprises establishing a database for storing reference inclination angle data under different wind speeds and blade lengths; when the measured inclination angle θ under the same wind speed after the blade is shortened deviates from the reference inclination angle in the database by more than a set threshold value, a tower drum fatigue warning is triggered.

[0046] By adopting the above-mentioned technology, compared with the prior art, the beneficial effects of the present application are as follows:

[0047] 1) The present application is different from the traditional contact type sensor, and utilizes the optical characteristics of the sea surface solar flare band to construct a dynamic horizontal reference, which can form a sub-pixel level stable reference line, improves the calculation precision of the angle between the tower drum axis and the reference line, and avoids the damage risk of the tower drum caused by salt spray corrosion and welding installation;

[0048] 2) The present application realizes a blade active protection mechanism: can adjust the blade length according to the real-time inclination angle change, and realizes the active excitation and online identification of the structural response by designing the blade as a “dynamic exciter”;

[0049] 3) The laser-vision coaxial detection design of the present application realizes precise positioning of faults. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 is the flow chart of the sea surface flare reference extraction and dynamic construction of the present application;

[0051] Figure 2 The tower inclination real-time measurement and blade protection control logic diagram of the application;

[0052] Figure 3 The monitoring-protection-diagnosis closed-loop control schematic diagram of the application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the application more clear and understandable, the application is further described in detail below in combination with the drawings and examples of the specification. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0054] On the contrary, the application covers any substitution, modification, equivalent method and solution made on the essence and scope of the application defined by the claims. Further, in order to make the public have a better understanding of the application, some specific details are described in detail in the following detailed description of the application. The application can also be completely understood without the description of these details by those skilled in the art.

[0055] The fan tower and blade adaptive protection system based on visual monitoring includes a visual monitoring device, a blade length adjusting unit arranged on the blade, and a fault response unit;

[0056] Specifically, the visual monitoring device includes a vibration analysis module and a salt mist-proof sealed cabin (IP68 protection) arranged at the bottom of the fan cabin. Among them:

[0057] (1) The salt mist-proof sealed cabin is coaxially integrated with:

[0058] 500 million pixel CMOS camera (frame rate 30 fps, focal length 50 mm, wavelength response range 400-1100 nm);

[0059] 532 nm pulse laser (maximum output energy 5 J / cm², spot diameter ≤20 mm, 10 Hz modulation);

[0060] Two-dimensional scanning galvanometer (positioning accuracy ±1 mm, scanning speed 100 points / s).

[0061] (2) Vibration analysis module (embedded module combined with hardware and software):

[0062] Hardware composition:

[0063] High-speed data acquisition card: receiving laser scanning vibration signal (10 Hz pulse);

[0064] Special processor (such as DSP / FPGA): real-time processing of vibration images transmitted by high-speed camera (1000 fps);

[0065] Storage chip: Solidified tower weld standard vibration waveform database.

[0066] Software Algorithm:

[0067] Digital image correlation (DIC): Analyzes the displacement field in images of vibration wave propagation;

[0068] Wave propagation delay model: Calculate the arrival time difference of vibration waves on both sides of the weld (threshold > 0.5ms indicates a fault);

[0069] Anomaly detection algorithm: Compare real-time waveforms with a standard waveform library to identify abnormal conduction delays.

[0070] Specifically, the blade length adjustment unit includes a retractable carbon fiber blade segment (1.5m in length, accounting for 10% of the total blade length), which includes an inner sliding cylinder (with attached carbon fiber blade) and an outer fixing cylinder. The outer fixing cylinder is inserted into the main blade and fixed on the main blade. The inner sliding cylinder is movably installed inside the outer fixing cylinder and is driven by a hydraulic actuator (output force 80kN, stroke accuracy ±0.5mm).

[0071] When the retraction condition is triggered (tower tilt angle θ > 0.8° for 60 seconds and wind speed ≤ 15m / s), the hydraulic actuator starts and outputs an axial thrust of 80kN. The hydraulic actuator pushes the inner sliding cylinder, causing it to slide inward within the outer fixed cylinder. The overall blade length shortens with the retraction of the inner sliding cylinder. During each retraction, the inner sliding cylinder only slides a distance corresponding to 5% of the blade length (the single retraction amount is controlled by the physical stroke). The cumulative retraction amount is limited by the structural design to ensure that it does not exceed 25% of the blade's rated length. The guide keyway fit accuracy, universal joint angle compensation, and the hydraulic actuator's ±0.5mm stroke accuracy together ensure the stability of the blade's physical structure during retraction and prevent sliding deviation.

[0072] Specifically, the functions of the fault response unit are as follows:

[0073] 1) When a loose point is located, the bolt pre-tightening device at the corresponding position is activated;

[0074] 2) When the vibration frequency deviation is >15%, start the cement grouting system.

[0075] The protection method for a visual monitoring-based adaptive protection system for wind turbine towers and blades includes the following steps:

[0076] See attached document Figure 1 As shown:

[0077] S1. Using a camera installed at the bottom of the wind turbine nacelle, continuously capture video including the tower and the sea surface at a pitch angle of 15°±3°;

[0078] S2, dynamically constructing the absolute horizontal datum of sea surface: detecting the solar flare belt profile of sea surface, fitting it into a smooth datum line after wavelet denoising, and fusing IMU data to compensate for fan swing; in the construction process, a mirror reflection light band with a wavelength of 500-600 nm is selected; a Kalman filter is used to suppress wave interference, and the filter cutoff frequency is set to 0.1H.

[0079] Referring to the drawings Figures 2-3

[0080] S3, extracting the edge lines of the tower drum on both sides, and calculating the average included angle between the edge lines and the horizontal datum line as the real-time inclination θ; the specific process is as follows:

[0081] S3.1, extracting the edge lines of the tower drum on both sides through an edge detection algorithm, and the specific process is as follows:

[0082] 3.1.1) input the original image (including the tower drum and the sea surface);

[0083] 3.1.2) use Gaussian filtering to suppress sea wave and salt spray noise in the original image;

[0084] 3.1.3) use the sobel operator to calculate the derivatives of the original image in X and Y directions respectively, and calculate the gradient intensity and direction of each pixel point based on the obtained derivatives;

[0085] 3.1.4) non-maximum suppression: traverse each pixel point, check the gradient direction of the two adjacent pixels, if the gradient intensity of the current point is greater than the intensity of the two adjacent points, keep the point as a candidate edge point; otherwise, set the suppression gray value of the current point to 0;

[0086] 3.1.5) set two threshold values: high threshold and low threshold, to calculate strong and weak edges, to distinguish between true edges and false edges caused by noise. Strong edge: gradient intensity > high threshold, weak edge: low threshold < gradient intensity < high threshold, when a weak edge point is connected with any strong edge point, and within the 8-neighborhood around the strong edge point (i.e. the strong edge point is located at the center of the nine-square grid, and the eight points around it), the weak edge point is kept as a true edge. Other unconnected weak edge points are discarded;

[0087] 3.1.6) after processing based on the edge detection algorithm of steps 3.1.1)-3.1.5), the original image is converted into a binary black and white image, wherein the white pixel points constitute all the edges, and the straight line segments are detected from the white points based on the Hough transform algorithm, and from all the detected straight lines, two straight lines that meet the tower drum edge characteristics are found, i.e. the edge lines.

[0088] S3.2, constructing a dynamic horizontal datum line, and the specific process is as follows:

[0089] ​The wave mirror reflection profile is extracted by using a wavelet denoising algorithm, and the wave interference is suppressed by a Kalman filter to form a sub-pixel level stable reference line; at the same time, the tower drum area is segmented by fusing LiDAR point cloud data, and the angle between the tower drum axis and the reference line is calculated in real time by using a Canny edge detection algorithm.

[0090] S3.3, calculate the inclination deviation , as follows:

[0091] The formula for calculating the inclination deviation of the tower drum is:

[0092] ;

[0093] wherein, is the inclination deviation of the tower drum, is the inclination of the tower drum measured in real time, is the theoretical inclination of the tower drum corresponding to the dynamic horizontal reference.

[0094] S3.4, based on the deep reinforcement learning model to optimize the output real-time inclination θ, the inclination adaptive prediction and control algorithm based on deep reinforcement learning is adopted, and the calculation formula is:

[0095] ;

[0096] ;

[0097] ;

[0098] wherein, represents the Q value of taking action in state ; is the learning rate, which is adaptively adjusted according to the inclination change to balance the stability and convergence speed of the algorithm; is the reward value at the current time; gamma is the discount factor; is the reference inclination change rate.

[0099] S3.5, update the system state in combination with the wind speed and the blade length.

[0100] S4, when θ is greater than the set angle threshold and lasts for t seconds, start the blade length adjusting unit to shorten the blade length in sections; the calculation formula of the blade length shortening amount in sections is:

[0101] ;

[0102] wherein, AL is the length of each blade shortening, K is the stiffness attenuation coefficient, which is obtained by exponential fitting of historical inclination data, θ is the real-time inclination, θ0 is the initial inclination, V is the current wind speed, and V0 is the rated wind speed. The above formula realizes dynamic optimization of the blade shortening amount by considering the influence of real-time inclination and wind speed on the blade shortening amount, and can reasonably adjust the blade shortening amount under different working conditions to better reduce the wind load of the tower.

[0103] In this embodiment, the blade protection control logic is as follows:

[0104] Trigger condition: inclination θ > 0.8° for 60 seconds.

[0105] The segmented contraction strategy is as follows:

[0106] .

[0107] Failure criterion: if the inclination change rate dθ / dt ≤ 0.01° / s after the nth contraction, immediately terminate the contraction and start the laser scanning.

[0108] The blade segmented shortening operation needs to meet the following conditions:

[0109] 1) Single shortening time is 10±2 seconds;

[0110] 2) The cumulative shortening amount does not exceed 25% of the rated length;

[0111] 3) When the wind speed exceeds 15 m / s during the shortening process, the operation is paused.

[0112] S5, monitor the inclination change rate after the blade shortening: if the change rate > 0.05° / s, it is determined that the elastic deformation is maintained, and the current blade length is maintained; if the change rate ≤ 0.01° / s, start the laser scanning diagnosis;

[0113] S6, move the laser focus of the pulsed laser along the tower weld seam through the scanning galvanometer control, and simultaneously analyze the vibration wave propagation image captured by the camera to locate the loose point coordinates. Wherein:

[0114] Dual-wavelength laser system is used for weld scanning during laser code scanning diagnosis, and the wavelength is dynamically adjusted according to real-time meteorological data: 532nm mode is used when visibility > 1km, and 1550nm mode is used when visibility ≤ 1km.

[0115] The switching response time is <0.5 seconds, ensuring the continuity of the vibration wave conduction time delay analysis.

[0116] When locating the loose point coordinates, the laser pulse frequency is 10Hz, and the spot diameter is ≤ 20mm; the propagation time difference of the vibration wave on both sides of the weld is compared, and the time difference > 0.5ms is marked as a fault point.

[0117] Specifically, the step S3 calculates the real-time inclination θ, and the step S4 establishes the inclination-airspeed-leaf length relationship database during the leaf segment shortening process, stores the reference inclination data under different airspeeds and leaf lengths; when the leaf is shortened, the inclination θ measured under the same airspeed deviates from the reference inclination in the database by more than 10%, the tower tube fatigue early warning is triggered.

[0118] The above merely describes preferred embodiments of the present application but should not be used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A visual monitoring based self adaptive protection system for wind turbine tower and blades characterized in that, The visual monitoring device, the blade length adjusting unit arranged on the blade, and the fault response unit are included. The visual monitoring device includes a vibration analysis module and a salt mist-proof sealed cabin arranged at the bottom of the fan cabin, the vibration analysis module is used for analyzing a displacement field of a vibration wave propagation image, calculating a vibration wave arrival time difference on both sides of a weld, comparing a real-time waveform with a standard waveform library, identifying a conduction delay anomaly, and the salt mist-proof sealed cabin is integrated with a camera, a pulse laser, and a scanning galvanometer. The blade length adjusting unit includes at least one telescopic carbon fiber blade segment for adjusting the length of the blade. The fault response unit is used for activating a bolt pre-tightening device at a corresponding position when locating a loosening point, and starting a cement grouting system when a vibration frequency offset is greater than a set threshold.

2. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 1, characterized in that, The method includes the following steps: S1, continuously shooting a video containing a tower drum and a sea surface by a camera installed at the bottom of the fan cabin at a set pitch angle; S2, dynamically constructing an absolute horizontal reference of the sea surface; S3, extracting edge lines on both sides of the tower drum, and calculating an average included angle between the edge lines and a horizontal reference line as a real-time inclination θ; S4, when θ is greater than a set angle threshold and lasts for t seconds, starting the blade length adjusting unit to segmentally shorten the length of the blade; S5, monitoring a change rate of the inclination after the blade is shortened: if the change rate is greater than a set threshold A, it is determined that the change is elastic deformation and the current length of the blade is maintained; if the change rate is less than or equal to a set threshold B, starting laser scanning diagnosis; S6, moving a laser focal point of the pulse laser along a weld of the tower drum by the scanning galvanometer, synchronously analyzing a vibration wave propagation image captured by the camera, and locating a loosening point coordinate.

3. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 2, characterized in that, The specific process of S2 is as follows: Detecting a solar flare belt profile of the sea surface, fitting a smooth reference line after wavelet denoising, and fusing IMU data to compensate for fan swing.

4. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 2, characterized in that, The specific process of S3 is as follows: S3.1, extracting edge lines on both sides of the tower drum by an edge detection algorithm; S3.2, constructing a dynamic horizontal reference line; S3.3, calculating the tilt deviation ; S3.4, optimizing an output real-time inclination θ based on a deep reinforcement learning model; S3.5, updating a system state in combination with a wind speed and a blade length.

5. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 4, characterized in that, The specific process of S3.1 is as follows: 3.1.1) inputting an original image; 3.1.2) suppressing sea wave and salt mist noise in the original image by using Gaussian filtering; 3.1.3) calculating a gradient intensity and a direction of each pixel point based on a derivative in X and Y directions of the original image calculated by using a sobel operator; 3.1.4) traversing each pixel point, checking two adjacent pixels in a gradient direction of the pixel point, if a gradient intensity of the pixel point is greater than intensities of the two adjacent pixels, retaining the pixel point as a candidate edge point; otherwise, setting a suppression gray value of the pixel point to 0; 3.1.5) calculating strong and weak edges to distinguish true edges from false edges caused by noise; 3.1.6) After processing based on steps 3.1.1) - 3.1.5), the original image is converted into a binary black and white image, where white pixels constitute all edges, and straight line segments are detected from the white points based on the Hough transform algorithm. From all the detected straight lines, two straight lines that meet the tower edge characteristics are found, which are the edge lines.

6. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 4, characterized in that, The specific process of S3.2 is as follows: A wavelet denoising algorithm is used to extract the sea wave mirror reflection profile, and a Kalman filter is used to suppress wave interference to form a sub-pixel level stable reference line. At the same time, the tower drum area is segmented by fusing LiDAR point cloud data, and the angle between the tower drum axis and the reference line is calculated in real time by the Canny edge detection algorithm.

7. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 4, characterized in that, In S3.3, the formula for calculating the tower inclination deviation is: ； wherein, is a tower tilt deviation, is a real-time measured tower tilt, is a theoretical tower tilt corresponding to a dynamic horizontal reference.

8. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to claim 4, characterized in that, In S4, the calculation formula of the blade subsection shortening amount is: ； Where ΔL is the length of each blade shortening, K is the stiffness attenuation coefficient, which is obtained by exponential fitting of historical inclination data, θ is the real-time inclination, θ0 is the initial inclination, V is the current wind speed, and V0 is the rated wind speed.

9. The method of protection of a visual monitoring based fan tower and blade adaptive protection system according to any of claims 2-8, characterized in that, It also includes establishing a database for storing reference inclination data under different wind speeds and blade lengths; when the blade is shortened, the measured inclination θ at the same wind speed deviates from the reference inclination in the database by more than a set threshold, triggering a tower fatigue warning.

Images